agentic-qe 3.6.16 → 3.6.18

package/.claude/agents/v3/qe-queen-coordinator.md

@@ -47,11 +47,15 @@ If unhealthy, report status and attempt recovery before proceeding.
 
 ```
 mcp__agentic-qe__memory_query({
-  pattern: "queen-orchestration-*",
-  namespace: "learning"
+  pattern: "queen orchestration patterns",
+  namespace: "learning",
+  semantic: true
 })
 ```
 
+Note: `semantic: true` uses HNSW vector search for natural language queries.
+For glob-style matching, omit `semantic` and use patterns like `"queen-orchestration-*"`.
+
 ### PHASE 4: Analyze Task and Select Domains
 
 Based on the user's request, determine which domains to activate:
@@ -245,8 +249,10 @@ mcp__agentic-qe__task_orchestrate({
 |------|---------|
 | `mcp__agentic-qe__memory_store` | Store patterns/learnings |
 | `mcp__agentic-qe__memory_retrieve` | Get stored data |
-| `mcp__agentic-qe__memory_query` | Search patterns |
+| `mcp__agentic-qe__memory_query` | Search patterns (supports `semantic: true` for HNSW vector search) |
 | `mcp__agentic-qe__memory_share` | Share between agents |
+| `mcp__agentic-qe__memory_usage` | Get memory usage statistics |
+| `mcp__agentic-qe__memory_delete` | Delete memory entry |
 
 ### Model Routing
 | Tool | Purpose |

package/.claude/skills/skills-manifest.json

@@ -904,7 +904,7 @@
   },
   "metadata": {
     "generatedBy": "Agentic QE Fleet",
-    "fleetVersion": "3.6.16",
+    "fleetVersion": "3.6.18",
     "manifestVersion": "1.3.0",
     "lastUpdated": "2026-02-04T00:00:00.000Z",
     "contributors": [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentic-qe",
-  "version": "3.6.16",
+  "version": "3.6.18",
   "description": "Agentic Quality Engineering V3 - Domain-Driven Design Architecture with 13 Bounded Contexts, O(log n) coverage analysis, ReasoningBank learning, 59 specialized QE agents, mathematical Coherence verification, deep Claude Flow integration",
   "main": "./v3/dist/index.js",
   "types": "./v3/dist/index.d.ts",

package/v3/CHANGELOG.md

@@ -5,6 +5,44 @@ All notable changes to Agentic QE will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.6.18] - 2026-02-21
+
+### Added
+
+- **6 new Agent Teams MCP tools** — `team_list`, `team_health`, `team_message`, `team_broadcast`, `team_scale`, `team_rebalance` expose ADR-064 domain team management via MCP
+- **Team membership in agent_list** — Each agent now shows its team role (`lead` or `teammate`) and team size
+- **Teams summary in fleet_status** — Fleet status response now includes active teams count, total agents in teams, and healthy team count
+- **Auto-team-wiring on agent spawn** — Agents are automatically registered in domain teams when spawned (first agent = lead, subsequent = teammates)
+- **addTeammate() on DomainTeamManager** — New method to add real agent IDs as teammates (instead of synthetic IDs from scaleTeam)
+- **Runtime message-type validation** — `team_message` and `team_broadcast` reject invalid message types with clear error messages
+
+### Fixed
+
+- **Race condition in parallel agent spawns** — Two concurrent spawns for the same domain no longer cause the second agent to become a ghost; the create-team failure now falls back to addTeammate
+- **Duplicate agent guard** — `addTeammate()` rejects agents already in the team, preventing corrupted team size calculations
+
+## [3.6.17] - 2026-02-21
+
+### Added
+
+- **HNSW semantic search** — `memory_query` now supports `semantic: true` for natural language queries using HNSW vector search with cosine similarity scoring, with automatic fallback to pattern matching
+- **Vector storage on write** — `memory_store` now indexes entries in the HNSW vector index alongside KV storage, so semantic search finds them immediately
+- **Hierarchical agent topology** — `fleet_init` with `topology: "hierarchical"` now assigns agents as domain leads or workers (first agent per domain becomes lead)
+- **MCP reconnection with buffering** — Transport errors trigger exponential backoff reconnection with request buffering and replay
+- **Real experience capture** — Tool executions now record patterns via ExperienceCaptureService instead of returning placeholder feedback
+
+### Fixed
+
+- **Stdio transport timeout under parallel load** — Write timeout increased from 30s to 120s with retry logic and backpressure handling, preventing dropped connections during parallel agent spawns
+- **Semantic search returned 0 results** — `memory_store` was only saving to KV store but not indexing vectors; now calls `storeVector()` after every write
+- **Embedding dimension mismatch** — Text embedding function now generates 768-dimension vectors matching the HNSW index configuration
+- **Transport stop() race condition** — `running` flag now set before closing readline, preventing error handler from triggering reconnect during explicit shutdown
+
+### Changed
+
+- Queen Coordinator documentation updated with semantic search parameters and complete tool reference table
+- Learning feedback messages now reflect actual capture status instead of hardcoded "patterns updated" text
+
 ## [3.6.16] - 2026-02-21
 
 ### Added

package/v3/assets/agents/v3/qe-queen-coordinator.md

@@ -47,11 +47,15 @@ If unhealthy, report status and attempt recovery before proceeding.
 
 ```
 mcp__agentic-qe__memory_query({
-  pattern: "queen-orchestration-*",
-  namespace: "learning"
+  pattern: "queen orchestration patterns",
+  namespace: "learning",
+  semantic: true
 })
 ```
 
+Note: `semantic: true` uses HNSW vector search for natural language queries.
+For glob-style matching, omit `semantic` and use patterns like `"queen-orchestration-*"`.
+
 ### PHASE 4: Analyze Task and Select Domains
 
 Based on the user's request, determine which domains to activate:
@@ -245,8 +249,10 @@ mcp__agentic-qe__task_orchestrate({
 |------|---------|
 | `mcp__agentic-qe__memory_store` | Store patterns/learnings |
 | `mcp__agentic-qe__memory_retrieve` | Get stored data |
-| `mcp__agentic-qe__memory_query` | Search patterns |
+| `mcp__agentic-qe__memory_query` | Search patterns (supports `semantic: true` for HNSW vector search) |
 | `mcp__agentic-qe__memory_share` | Share between agents |
+| `mcp__agentic-qe__memory_usage` | Get memory usage statistics |
+| `mcp__agentic-qe__memory_delete` | Delete memory entry |
 
 ### Model Routing
 | Tool | Purpose |